Supermicro Starts Shipping NVIDIA Vera Rubin NVL72 Rack Systems

US server maker Supermicro says it has started shipping NVIDIA Vera Rubin NVL72 rack systems. The company built a full Data Center Building Block Solution (DCBBS) around it, including the liquid-cooled compute rack itself, a 1.8MW in-row cooling distribution unit (CDU), optional rear-door heat exchangers, and network integration and cabling services based on NVIDIA reference architecture.

The DCBBS blueprint supports power ranges from 5MW up to gigawatt scale. A single scalable unit blueprint covers 16 compute racks, providing 1,152 Rubin GPUs and 331TB of HBM4 memory, along with supporting cooling, power, storage, and networking infrastructure.

This marks the start of large-scale delivery for NVIDIA's next-generation Rubin architecture AI infrastructure, offering a complete rack-level solution for hyperscale data centers deploying next-gen AI training and inference clusters.